Performance Evaluation of Face Recognition for Access Control System

Deep learning-based facial recognition system in real-world circumstances of repeated authentication may result in mismatch due to inconsistencies in face image conditions that are incompatible with database calls. Face recognition performance is influenced by environmental factors such as light and local background images. As a result, this paper investigates the effects and causes of failure of facial recognition system. The HIKVISION model and the algorithm of DS-K5671- 3XF/Z from the DS-K1T671T series were tested and studied. In the experiment, it was discovered that irregularities in testing faces and face image circumstances, as well as light intensity, were the most relevant factors affecting the facial recognition system. Minor changes of the certain factors may degrade the performance significantly